Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SSDF ratings are 100% accurate

Author: Ed Schröder

Date: 02:42:55 12/13/99

Go up one level in this thread


>Subject: Re: SSDF ratings are 100% accurate

>Posted by John Warfield on December 12, 1999 at 20:36:17:

>>>Now I am beginning to see that SSDF ratings do not reflect performance against
>>>humans ­ period.  Going back to my example, program B could actually be weaker
>>>than program A against GMs, even though it is 50 points stronger in SSDF comp
>>>vs. com testing.
>>>
>>>I guess this is what Ed Schroder has been saying all along about Rebel.  I
>>>need to think about this for a while.
>>
>>The MAIN difference I noticed: in comp-comp both programs (in many cases)
>>can afford (multiple) small to big positional mistakes. Try this against
>>a GM, one little mistake and you lose. The REBEL-HOFFMAN game was a
>>perfect example of this.
>>
>>Ed
>
>
>  The rebel hoffman game was lost because rebel crashed remember?

Of course I remember. But in the meantime I have changed my mind. Rebel due
to the hardware problems used about 50-60% of its time also 2 horrible moves
were played (one with a +2.xx) score that couldn't be reproduced. This made me
decide the game was worthless.

But after going through the game again and again my conclusion is different now.
Rebel didn't understand the opening, played a few inferior moves and technically
the game was over after move 18. Note that Rebel was in book till move 15!

I think that's all there is to say about this game. The hardware problems came
after the 2 inferior moves (16.Qc1 and 18.a4) and GM Hoffman did not let Rebel
go. In comp-comp however you still would have good chances to win the game
(note that after 18.a4 Rebel is still a pawn up) as the opening was very
strategic by nature an area computers are still weak.

Just try any chess program that gives you a positive score for black after 18.a4
and if it does buy it by all means :-)

Or take 2 (or more) good chess programs and let them continue after 18.a4 and
I am pretty sure white's total game score will be over 50%.

This is what I got as a comment from GM Hoffman about the game:

[ begin ]

I think it was a very interesting game for black,with 15... Rb7 an interesting
novelty.16. Qc1 means that Rebel doesn't understand the position (16.Rc1
was normal plan).

I think it is very hard to a computer to know the difference between to have
material plus and the strategical compensation for the pawn. That's because
I choice the Volga Gambit. That you must think how to improve for a high level
program.

[ end ]

Rebel was caught on a weak point of its opening book. Very clever and an
instructive experience.

Ed





This page took 0.01 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.